Functional Neuroimaging in High-Risk 6-Month-Old Infants Predicts Later Autism

Background: Effective early interventions that can ameliorate the defining deficits of ASD and improve long-term outcomes rely on early detection. One barrier to early (i.e., prior to 24 months) detection is that the defining behavioral characteristics of ASD generally unfold during the second year of life, sometimes not showing consolidation of the full behavioral syndrome around 24 months or later.

Recent research using functional connectivity magnetic resonance imaging (fcMRI) has linked the functional organization of the human brain to individual cognitive profiles. These measures of brain functional connectivity are reliable and can accommodate participants as young as neonates. Furthermore, in conjunction with machine learning approaches, fcMRI data has provided predictions of brain maturation and diagnostic category at the single-subject level.

Objectives: We aimed to use functional neuroimaging with 6-month-old infants to identify which individual children will receive a research clinical best estimate diagnosis of ASD at 24 months of age.

Methods: Prospective neuroimaging and behavioral data were collected from 59 naturally sleeping infants at high familial risk for developing ASD. First, we defined functional connections in the 6-month-old brain that correlated with 24-month scores on assessments of social interactions, communication, motor development, and repetitive behavior – all features common to the diagnosis of ASD. We then used a fully cross-validated machine learning algorithm to demonstrate that fcMRI can identify 6-month-old infants who progressed to a clinical diagnosis of ASD at 24 months of age. For each infant, we defined brain features and trained a classifier using an independent set of 58 high-risk infants. We then predicted that infant’s future diagnosis using only information from their 6-month functional neuroimaging scan. To test the generalizability and validity of our results, we used a similar classification analysis with a greater number of subjects held independent (leave-10-out).

Results:  The overall classification accuracy for later ASD using functional connectivity data in 6-month old infants was 96.6% (95% CI 87.3–99.4, p<0.001). The positive predictive value of this approach was 100% (95% confidence interval [CI], 62.9–100), correctly predicting 9 of 11 infants who received a diagnosis of ASD at 24 months (sensitivity 81.8% [95% CI 47.8–96.8]). All 48 6-month-old infants who were not diagnosed with ASD were correctly classified (specificity 100% [95% CI, 90.8–100]; negative predictive value 96.0% [95% CI 85.1–99.3]). On average, the leave-ten-out analysis performed with 92.7 ± 0.7% accuracy.

Conclusions: These findings demonstrate the potential for early detection of autism in infants at high familial risk and serve as a proof-of-concept that patterns of infant brain measures precede the defining behavioral characteristics of ASD. Ultimately, this study represents an initial, but critical, first step toward developing infant diagnostic methods and enabling efficient tests of infant interventions.